Effective cosmic density field reconstruction with convolutional neural network
Authors:
Xinyi Chen,
Fangzhou Zhu,
Sasha Gaines,
Nikhil Padmanabhan
Abstract:
We present a cosmic density field reconstruction method that augments the traditional reconstruction algorithms with a convolutional neural network (CNN). Following Shallue $\&$ Eisenstein (2022), the key component of our method is to use the $\textit{reconstructed}$ density field as the input to the neural network. We extend this previous work by exploring how the performance of these reconstruct…
▽ More
We present a cosmic density field reconstruction method that augments the traditional reconstruction algorithms with a convolutional neural network (CNN). Following Shallue $\&$ Eisenstein (2022), the key component of our method is to use the $\textit{reconstructed}$ density field as the input to the neural network. We extend this previous work by exploring how the performance of these reconstruction ideas depends on the input reconstruction algorithm, the reconstruction parameters, and the shot noise of the density field, as well as the robustness of the method. We build an eight-layer CNN and train the network with reconstructed density fields computed from the Quijote suite of simulations. The reconstructed density fields are generated by both the standard algorithm and a new iterative algorithm. In real space at $z=0$, we find that the reconstructed field is $90\%$ correlated with the true initial density out to $k\sim 0.5 h{\rm Mpc}^{-1}$, a significant improvement over $k\sim 0.2 h{\rm Mpc}^{-1}$ achieved by the input reconstruction algorithms. We find similar improvements in redshift space, including an improved removal of redshift space distortions at small scales. We also find that the method is robust across changes in cosmology. Additionally, the CNN removes much of the variance from the choice of different reconstruction algorithms and reconstruction parameters. However, the effectiveness decreases with increasing shot noise, suggesting that such an approach is best suited to high density samples. This work highlights the additional information in the density field beyond linear scales as well as the power of complementing traditional analysis approaches with machine learning techniques.
△ Less
Submitted 18 June, 2023;
originally announced June 2023.
Range contraction enables harvesting to extinction
Authors:
Matthew G. Burgess,
Christopher Costello,
Alexa Fredston-Hermann,
Malin L. Pinsky,
Steven D. Gaines,
David Tilman,
Stephen Polasky
Abstract:
Economic incentives to harvest a species usually diminish as its abundance declines, because harvest costs increase. This prevents harvesting to extinction. A known exception can occur if consumer demand causes a declining species' harvest price to rise faster than costs. This threat may affect rare and valuable species, such as large land mammals, sturgeons, and bluefin tunas. We analyze a simila…
▽ More
Economic incentives to harvest a species usually diminish as its abundance declines, because harvest costs increase. This prevents harvesting to extinction. A known exception can occur if consumer demand causes a declining species' harvest price to rise faster than costs. This threat may affect rare and valuable species, such as large land mammals, sturgeons, and bluefin tunas. We analyze a similar but underappreciated threat, which arises when the geographic area (range) occupied by a species contracts as its abundance declines. Range contractions maintain the local densities of declining populations, which facilitates harvesting to extinction by preventing abundance declines from causing harvest costs to rise. Factors causing such range contractions include schooling, herding, or flocking behaviors--which, ironically, can be predator-avoidance adaptations; patchy environments; habitat loss; and climate change. We use a simple model to identify combinations of range contractions and price increases capable of causing extinction from profitable overharvesting, and we compare these to an empirical review. We find that some aquatic species that school or forage in patchy environments experience sufficiently severe range contractions as they decline to allow profitable harvesting to extinction even with little or no price increase; and some high-value declining aquatic species experience severe price increases. For terrestrial species, the data needed to evaluate our theory are scarce, but available evidence suggests that extinction-enabling range contractions may be common among declining mammals and birds. Thus, factors causing range contraction as abundance declines may pose unexpectedly large extinction risks to harvested species.
△ Less
Submitted 29 March, 2017; v1 submitted 1 October, 2014;
originally announced October 2014.
